Do Election Forecasts Suffer from a Lack of Economics?

Forecasts of this election are certainly missing economics.  However, lacking knowledge and skills of the other tools and methods for election forecasting, I cannot say how much the absence of economics matters.  My only purpose here is to cite the missing economics and attempt to assess the direction of the forecast bias as it pertains to the Presidential election.  The missing economics fit into three categories: information aggregation, voter incentives to tell the truth, and voter incentives to participate in the general election.

 

At the time of writing (Oct 30), the RealClearPolitics Poll Average is Biden +7.9 (51.4-43.5) nationally and +3.1 (48.9-45.9) in the top battleground states.[1]  Based on a very similar dataset, FiveThirtyEight puts Biden’s chance of winning at 90 percent while The Economist puts it at 96 percent.  FiveThirtyEight (Economist) forecasts that Biden will win by 156 (162) electoral votes, respectively.

 

The polls used for these averages and forecasts (hereafter, “the forecasters”) contact potential voters by landline, cell phone, and internet and ask them “Who do you think you will vote for in the election for President?” (emphasis added).  Other questions are asked about the election, but it does not appear that the answers to these questions are part of the aforementioned averages or forecasts.

 

 

Information Aggregation

 

The information aggregation methods used by forecasters seem inferior to alternatives that have been well-studied, readily available, and better reflect results on the economics of information.  The methods used by forecasters are often more than ten percentage points more favorable to Biden than the alternatives.  I have not seen any explanation from the forecasters why the alternatives are not part of their forecasts.

 

U.S. voters are heterogeneous and scattered over millions of square miles.  One of the challenges for an election forecaster is to assemble or reweight a polling sample that is representative of those who actually vote.  For this purpose, the polls used in the averages and forecasts do some reweighting of their samples by demographics.  Most famously, less-educated voters are often (but not always) upweighted given that their relative propensity to vote in recent years has exceeded their relative propensity to participate in polls.  Other demographic variables sometimes used for these purposes are voter registration and voting history.

 

As recently as a few years ago, it was well known that another poll question has a much better track record at predicting election outcomes: “Who do you think will be elected President in November?”  This “expectation” question is included in some of the polls used by the forecasters but, as far as I can tell, are not any part of the forecast.[2]

 

David Rothschild and Justin Wolfers found in 2012 that the optimal forecasting weight on the expectations question is about nine times the weight on the own vote question.  Furthermore, the optimal weight on voter expectations is, they said, even greater when: (1) “voters are embedded in heterogeneous ... social networks,” (2) “they don’t rely too much on common information,” and (3) “small samples are involved.”  Two of the three of these criteria describe 2020 voters even better than they described voters in the past.

 

A couple of pollsters express interest in voter expectations and find them to deviate significantly from the own vote intentions that are the foundation of election forecasts.  The USC Dornsife poll (7-day window ending October 29) finds Biden leading by 11.6 in own vote intentions but only 2.1 points in election expectations!  When respondents are asked about how they expect their social contacts to vote, Biden leads by 5.3 points.[3]  The Fox News Poll found Biden winning by 10 points on the own-vote question but losing by 9 points on “do you think more of your neighbors are voting for Joe Biden or Donald Trump?”

 

In other words, the own-vote question that is the foundation for the forecasts differs from the other questions by 6 to 19 percentage points, which is several multiples of the RCP battleground average Biden-Trump gap of 3 percentage points.

 

Gallup, which does not even ask the own vote intention question, finds that 56 percent of Americans expect Trump to win.  Betting markets put Trump’s chances at about 34 percent.  Both of these are far different from the election forecasts built on own-vote polls.

 

Betting markets are rich enough to allow bets on the electoral margin of victory.  The most expensive contracts are Biden 150-209 and Trump 60-99.  One of these coincides with the election forecasts but the other is wildly different.  In other words, betting markets seem to put significant probability on the event that the polls used by the forecasters are way off in Biden’s favor.[4]

 

 

Voter Incentives 1: Social Desirability

 

The incentives to truthfully participate in polls are different from those to participate in the general election and these differences are correlated with political affiliation.

 

One of the potential incentives is “social desirability.”  A Cato poll found that 62 percent of Americans “say the political climate these days prevents them from saying things they believe because others might find them offensive.”  Another study found that 11.7 percent of Republicans said they were reluctant to tell the truth on telephone polls (about the Presidential race), as compared to 5.4 percent of Democrats and 10.5 percent of Republicans.

 

Trump is the Bad Orange Man to many.  Biden is the overwhelming choice among the rich and famous, many of whom blamed Trump’s 2016 victory on his allegedly deplorable racist supporters.  More recently, property destruction (looting) has become more common and may be related to anti-racism in the minds of some voters.  This suggests that some fraction of Trump supporters would not acknowledge their support for him – the “shy Trump voter” – especially those in Democratic communities.

 

The quantitative question of course is the magnitude of the effect of shy-Trump and shy-Biden voters on poll results relative to election results.  I have been amazed how little effort the forecasters put into assessing this magnitude and applying a correction.  Today Nate Silver dismissed the shy Trump voter theory by transforming it into a straw man that a huge red wave is coming, which he says is contradicted by early voting data.  In the past he did nothing more than take the difference between the results of phone polls and internet polls (note that today Biden’s edge is two percentage points more in phone surveys, which is not negligible and is in the expected direction).[5]

 

Social desirability bias is one of the reasons why pollsters have fielded the voter-expectation and voter-neighbors questions.  Mr. Silver says this approach is “stupid” because it deviates from “the shy-Trump voter narrative.”

 

Another step, taken by the Democracy Institute, Trafalgar and others, is to repeatedly assure respondents that their responses are fully confidential.  They assert that the assurances affect the poll results, although I have not seen estimates of the magnitude.

 

 

Voter Incentives 2: Turnout

 

Voting in person or by mail is different than picking up the phone or opening an email to begin an online survey.  The difference is particularly stark during a pandemic.  I do not see that forecasters are accounting for any relationship between the difference and political affiliation.

 

Republicans are doing much less to withdraw from normal activities in an effort to protect themselves from COVID-19.  Gallup found that more than 77 percent of Democrats worried about getting the coronavirus whereas less than 29 percent of Republicans worried about that.  More than 70 percent of Democrats were avoiding going to public places as compared to less than 38 percent of Republicans.[6]  Meanwhile, confirmed cases are surging in the Midwest.  A widely televised expert even went so far as to say that Midwesterners would not be adequately protected by a cloth mask and that going out in public required a N-95 mask instead.

 

In short, COVID-19 surges create a perceived cost of in-person voting that is likely greater for Democrats than Republicans.  Some will switch to voting by mail but perhaps others will not vote at all.  I do not have an estimate of the magnitude of this effect.

 

Overall, economics by itself suggests that the widely cited polls are exaggerating the electoral advantage, if any, that Biden will have on election day.  I am no polling expert and therefore cannot assess the magnitude of these polling biases.  Betting markets and other methods of information aggregation do not show as much optimism for Biden as do the election forecasters.  These markets appear to put significant probability mass on the event that Biden’s election results are far worse than his polling results. 

 

 

---

[1] FL, PA, MI, WI, NC, AZ.

[2] E.g., FiveThirtyEight.com provides the data used for its forecasting model and “expectation” questions are not listed therein.

[3] My purpose of citing USC is not to assert that it is more accurate than others, but merely because it fields the various questions of interest in the same samples so that sample procedures can be held constant.

[4] Josh Barro provides some reasons why the betting markets might be skewed.  I have myself seen election betting markets depart wildly from objective reality (traders apparently unaware of what exactly they were trading).  With that said, a profit maximizing bookmaker might subsidize or discount election betting in order to draw in customers who would continue betting with their winnings after the conclusion of the election.  If Trump voters are especially elastic to the discounts, bookmakers may price Trump contracts low (perhaps a couple of percentage points).

[5] Mr. Silver has seen a lot of data related to elections, which may have revealed to him that shy Trump voters are negligible for reasons that he cannot (or will not) articulate in public.

[6] The study was August 2 through September 27.

Unions and Inequality: Looking at the Obvious

At the same time that a particular labor union (teachers) has successfully pushed to keep school buildings closed, we are reminded that unions are part of the "solution to inequality."  As an update to an old literature on unions and inequality, let's first look at distributional effects of closing school buildings.

The chart shows that closing school buildings reduces learning for all groups, but especially low-income and minority pupils.  So the actions of teacher unions today (they are far more likely to have their schools closed) will add to inequality in the future as the pupils enter the labor market.

Now let's look at today's labor market.  The weekly cash earnings of teachers are 22 percent more than those of nonteachers.  As a result of being offered more and richer fringe benefits (see the table below), their weekly compensation is 39 percent more, which I estimate assuming that retirement benefits involved an average 5 percent employer contribution for nonteachers and value teacher's pensions according to actuarial calculations (17 percent of salary).

It appears that an occupation that is well above average income is closing schooling buildings to the detriment of millions of low-income and minority children.

Are Regulations "Job Killing"?

The traditional models of regulations and growth treat regulation as an adverse productivity shock (more inputs for the same output) in order to help the environment, fairness, or some other social good.  But a productivity shock has opposing income and substitution effects on labor supply.  Arguably a regulation that works as a productivity shock has no aggregate effect on jobs.

Reminded how Gary Becker many times told me that "somebody benefits," I do not endorse the productivity-shock model of regulation, at least as relates to the Federal regulations added and removed over the past 20 years.  In economics jargon, the "rectangle" created in a market by regulation is not entirely wasted: some of it is a transfer to special interests and therefore not an income effect in the aggregate.  This kind of regulation is more like an excise tax with the revenue paid to special interests.  Excises taxes unambiguously reduce aggregate equilibrium employment.

As the CEA showed in the 2019 and 2020 Economic Reports of the President, many of the regulations removed by the Trump Administration were more like the excise tax than like a productivity shock.  That's why special interests fought back so hard (and a couple of times, they won).  

[Even a productivity shock reduces employment in the short run to the extent that it reduces the productivity of investment.  i.e, that's another way that the "job killing" can occur.  There is another interesting case in which the rectangle from regulation is a transfer from Americans to foreigners in which case the employment reduction is outside the country.]

The sign of the effect on the employment of the regulated industry is also ambiguous.  If the regulation is a transfer from consumers to producers, with no adverse productivity effect, the regulation will reduce industry employment because that transfer is achieved by restraining supply.  But that frees up resources for other industries, which is why the aggregate employment effect can be nil.

To the extent that regulation reduces productivity in the regulated industry, we need Marshall's Laws of Derived Demand to sign the effect on industry employment.

One application provided in Chicago Price Theory is the regulation of illegal drugs.  Their demand is price inelastic in the sense that drug prohibition reduces consumption (although see here for a tragic exception) while it increases what consumers spend on drugs.  The price elasticity of demand is an important part of Marshall's Laws.  For illegal drugs, the result is more people working to (or serving prison time for) grow, manufacture and distribute illegal drugs because law enforcement reduces their "efficiency."  But those people are coming out of other activities, which is why the aggregate employment effect can still be nil. 

Shoddy Executive Order Bears Fingerprints of Navarro and Krugman

An immigration Executive Order was issued two days ago.  I read it yesterday and gathered my thoughts and relevant memories over the subsequent 24 hours.

The EO contains immigration regulations and purported economic justifications for the regulations.

The EO’s economic justification is essentially that it is good to suppress labor supply during a recession.  I disagreed with such a conclusion when it was offered years ago by Krugman, Eggertson, and others.  The conclusion is just as wrong when it comes from President Donald Trump.

The empirical fact, which is not a surprise from a theoretical point of view, is that labor supply and demand matter just as much at the margin during a recession as they do during an expansion.  See Chapter 8 “Recession-era Effects of Factor Supply and Demand” of my 2012 book (a more recent JPE paper confirmed these results but I cannot find the link right now).  A recession is not an economic excuse for suppressing labor supply.

The faulty economic analysis I see in the EO sounds to me like Peter Navarro talking.  Hearing his voice now catches me a bit by surprise because, although he is a part of the populist story, his “rudeness, ignorance, and dishonesty” are well known from the President on down.  [I believe that Hassett, Mnuchin, Mulvaney, and Kudlow shot down such Navarro initiatives in the past, although I was not present at those meetings or even much involved with the prep.]

Suppressing labor supply is also poor public relations.  The employment and productivity numbers will come in lower than they would with a more market-oriented recovery.  (Only a couple of the Navarro stories were included in my 2020 book because they were a small fraction of my experience, and the President is a lot more interesting.  But a hilarious one – in the reader-spits-out-coffee category – is about another time that Navarro flunked marketing.)

The justification for the EO’s regulations, if there is any, would have to be that it somehow begins a path to fixing the broken status quo system that was in place before Monday.  That system was full of special-interest favors, which the President should be removing as he has removed them in many other regulatory areas.  I am pessimistic (i.e., optimistic for the entrenched special interests) that there is any such path in the immigration area, though.

Gary Becker’s immigration plan should be given serious consideration.  President Trump agrees with that on purely economic grounds (Chapter 6 of my book), although he sees that as a political nonstarter (“radical” as Becker put it).  Perhaps the immigration plan Trump proposed in May 2019 (essentially the Canadian and Australian systems) is a more politically correct approximation to the Becker plan.

Part of this EO pertains to foreign-born scholars working in the U.S., which saddens me personally.  My closest friends are squarely in that category.  The value they add is so great that policy will likely change so that they continue to work in the U.S. 

A little known fact is that President Trump is a very good listener (my book is filled with examples; that’s how he became a populist president).  So speak up!  He may decide in your favor.  After listening, he may articulate your position better you do.  In that case, I’m sorry because that is a strong indication that he is deciding against you (e.g., here) and wants you to at least know that you were heard.

Dueling Memoirs: Mulligan vs. Bolton

What do readers have to say?

Style

A New York Times Review says that Bolton's memoir "has been written with so little discernible attention to style and narrative form...."

One reader (who asked to remain anonymous because he/she fears retribution at work) finds You're Hired! to be an "extremely well-written book." Brian Blase found it to be "enjoyable and easy to read."

Tone

Back to the review of Bolton, "Underneath it all courses a festering obsession with his enemies ...the book is bloated with self-importance."

Pages xii and xviii of You're Hired! explain how I was "the Apprentice" and that readers should be "prepared to be as amazed and humbled as I was."

By Joe Grogan's reading, You're Hired! is an "an insightful, honest, book ...  free from score settling and self promotion."

Substance

You're Hired! shows that Bolton has distorted the truth, omitted important context, and contradicts critical facts in plain sight.  Excepts related to this matter were published on-line as Bolton is Wrong; I was There.

Coronavirus and What it is Like to Speak with Trump

One viewer says "You play it very straight which at this point is unheard of in Washington."

The Higher Ed Market after Floyd

@GlennLoury objects to what university administrators are doing. They are economic actors too, who will not benefit from a repeat of 1968, so it is predictable that their reactions might risk some scholarship, reason, and learning.
But there is also competition in the industry and thereby an opportunity for an (aspiring?) administrator who expresses the interests of the many individuals who have not yet reached the fashionable conclusions. Something like the famous Zimmer letter.
This competition might play out slowly given that (barring regulatory favors) universities will now compete in another important dimension: whether 2020-21 students are allowed to purchase a college education that does more than Zoom (which would have made 1968 impossible).

Moreover the Zimmer letter was not written on a blank tablet. A Uchicago committee led by @stone_geoffrey had already worked on it 2 years before. If this capital does not yet exist for the current situation, it will take time to build it up.

Labor Market Recovery Begins when States Begin to Open

The first chart below is an estimate of weekly US employment per adult.  It suggests that the bottom was the week ending May 7, and that a recovery may have begun.

The estimated recovery may not look large on the scale of the current depression, but it is about 7.5 million employees above May 7 and 3 million employees above late April.  Note that the entire recovery from the 2008-9 recession was "only" 7 million employees above population growth and took ten years rather than a week or two.

At about the same time, states began ending their stay-at-home orders.  E.g., Texas May 1 and California May 8.  I expect another increase in early June as more reopening occurs.  A big increase will occur when UI bonuses expire, which may be as early as August.

The imputation is based on the scatterplot below.

Measuring Employment between Monthly Surveys

The Employment Situation Report by the Bureau of Labor Statistics comes only monthly.  It measures only the seven-day week (or, with the establishment survey, pay period) including the 12^th of the prior month, which means that this month four very interesting weeks will be skipped and that the report on that April week will not be released until May 8.

Three data sources provide employment information on at least one of the missing four weeks, with the results shown in the chart below.  The results suggest that employment has fallen more than 20 million and perhaps as much as 36 million by April 11.  This does not begin to count employees who had their hours reduced.

One is an attempt by Bick and Blandin (2020) to imitate the BLS household survey for the week of Sunday March 29 to Saturday April 4.  They sampled 1,118 respondents, finding that employment per adult aged 18-64 was 17.8 percent below what it was in February and 17.4 percent below what it was in January.  They find that hours worked per person (including zeros for those not employed) were 27.7 percent below what they were in January and February.

Coibion, Gorodnichenko, and Weber (2020) surveyed 18,344 members of the Nielsen HomeScan panel during the days Thursday April 2 through Monday April 6.  The respondents were asked “Do you have a paid job?” which is different from the BLS questions but the same as January surveys of the Nielsen panel.  Their sample shows a 12.5 percent decline from January to April.

As a third source, I use the excess of continued UI claims for the week ending April 4 rescaled by 0.4, which is the typical ratio of continued claims to persons unemployed during the 2008-9 recession.[1]  The rescaled amount is 16.1 percent of February employment as measured by the February household survey.  This approach also offers employment estimates for the weeks before and after the week ending April 4.

Especially during the pandemic, “employment” and “unemployment” can vary significantly merely due to definitions.  Is a person on the payroll but told not to work considered employed?  The BLS knows from its experience with Federal shutdowns that surveyors and respondents frequently misclassify relative to the technical definition in the survey.  The practical classification grey areas are also presumably sensitive to question wording.  UI claims also have a grey area that presumably changes as new Federal policies increase the financial reward to unemployed rather than out of the labor force.

These measurement challenges suggest using hours worked rather than employment and using multiple data sources, which are not entirely congruent approaches because only one of the three sources measures hours worked.[2]  I therefore measure the decline in hours worked by averaging the three employment estimates and then applying the Bick-Blandin estimate of the decline in hours per employee.

Regarding initial claims versus continued claims, initial claims may not be granted due to ineligibility and do not show a stable ratio to employment changes during the 2008-9 recession.  Initial claims are reported a week ahead of continued claims.[3]

---

[1] The average continued claims was 1.7 million in both January and February, which is the baseline from which I calculate the “excess.”

[2] Hours worked are also of interest because many people were under employed in April (Bick and Blandin 2020).

[3] In March and April 2020, continued claims may include an abnormal share of ineligible claims due to abnormal delays in state processing, although this effect should disappear over a horizon long enough for states to process the claims.  On the other hand, the CARES Act passed March 27 will begin distorting the relationship between employment and continued claims because the Act included a large UI bonus that will encourage an abnormally large fraction of the eligible unemployed to apply.